Interpretable AI

Turning data into trusted action

With proprietary AI technologies, we build business solutions that simultaneously deliver full explainability and state-of-the-art performance.

Bringing interpretability to machine learning and artificial intelligence

Our core technologies have been created from the ground-up to simultaneously deliver interpretability and state-of-the-art performance. Our focus on interpretability means:

  • Our technologies are effective and readily usable in applications that require interpretability (e.g. banking regulations on fair lending).
  • Model creation is transparent and easy to debug, resulting in a more efficient development process and a shorter time to impact.
  • Machine learning and AI solutions can be easily explained and understood by all stakeholders – from data scientists to executives – making them easy to deploy with confidence and effect change.
Optimal Decision Trees

Interpretable predictive modeling with performance comparable to random forests or gradient boosting

Optimal Imputation

Understandable processing of unclean data that leads to superior predictive models

Optimal Feature Selection

Simplify the data science workflow with single-step, repeatable and performant feature engineering and selection

From Predictions to Prescriptions

Data-driven decision making with state-of-the-art optimization

End-to-end Solutions

Built on the foundation of our core technologies, we deliver interpretable and performant end-to-end solutions for business problems. Some examples of these solutions include:


Interpretable learning of retail banking customer behaviors and goals from transactional information


Best-in-class automatic real-time malware detection algorithms

Health Care

Medically-validated tools for predicting surgical outcomes and cancer mortality


Automated claims processing and insurance claimant segmentation

Our Team

Dimitris Bertsimas, Partner

Dimitris is a Co-Founding Partner of Interpretable AI and the Co-Director of the Operations Research Center at MIT. He has received numerous research awards, has written over 200 research papers and 4 graduate textbooks that are used around the world. He was a Co-Founder of Dynamic Ideas LLC, the assets of which were sold to American Express in 2002.

Jack Dunn, Partner

Jack is a Co-Founding Partner of Interpretable AI. He has developed many novel analytics approaches including the Optimal Trees methodology, and has considerable experience applying machine learning and AI to problems in both research and industry settings. He has a PhD in Operations Research from MIT.

Daisy Zhuo, Partner

Daisy is a Co-Founding Partner of Interpretable AI. She has expertise in developing scalable machine learning techniques including Optimal Imputations, with extensive research and industry experience in applications of analytics and AI systems in health care. She has a PhD in Operations Research from MIT.

Jeremy Toledano

Jeremy is a Research Scientist at Interpretable AI. He is passionate about applying state-of-the-art technology to real-world problems and has deep expertise in natural language processing. He holds a Master of Business Analytics from MIT and a MS in Applied Mathematics from École Centrale Paris.

Maxime Amram

Maxime is a Research Scientist at Interpretable AI. He leverages his expertise in machine learning to drive value for organizations, with extensive experience in the latest deep learning developments. He studied data science during his master's at MIT and holds a MS in Applied Mathematics from École Centrale Paris.